Nonlinear Prediction of Short-Time Sea Clutter
Qiao Xiang-wei · Radar Science and Technology · 2009
Sea clutter prediction is the key point in the research on radar signal processing and target detection.As the fact that sea clutter has chaotic characteristic and is nonlinear and nonstationary.Normalized RBF neural network and least squares support vector machines(LSSVM) are applied to nonlinear prediction of sea clutter time series.Considering sea clutter being the backscatter of the moving sea surface,so the spatial information should be considered for prediction.A new method of LSSVM-CML(Coupled Map Lattice) is presented for sea clutter spatiotemporal prediction,which has more physical meaning.Real sea clutter data is used as initial data and comparison for prediction,mean square error(MSE) and maximum absolute error(MAE) are derived as the evaluation criteria of prediction precision.Experiment results indicate that since the consideration of sea clutter spatiotemporal information,LSSVM-CML has the best prediction precision.